On datasets: initially used revdotcom/earnings22 dataset but due issues with the dataset (no dataset card, data being absent), a decision was made to switch to distil-whisper/earnings22
Came across distil-whisper models which are faster, lighter, and supposely should be tested